Startup Performance

transport::tests::startup_budgets guards the work controlled by this SDK with a deterministic local JSON-RPC fixture. Each metric uses five warmup iterations and 50 measured samples. The test reports median and p95 latency and fails if any p95 reaches 50 ms.

• publicImportMs: the first public autohand_sdk::initialize() call timed inside a fresh Rust test process, excluding Rust runtime process boot. This idempotent hook performs the crate's eager runtime initialization.
• sdkStartReturnMs: elapsed time for the public AutohandSdk::start API, including its readiness getState request.
• fixtureSpawnToFirstRpcMs: elapsed time to spawn the deterministic fixture and complete a successful getState through the transport.

Baseline captured on 2026-07-20:

Metric	Median	p95	Budget
publicImportMs	0.003417 ms	0.006750 ms	< 50 ms
sdkStartReturnMs	5.668250 ms	6.774875 ms	< 50 ms
fixtureSpawnToFirstRpcMs	5.564083 ms	6.308125 ms	< 50 ms

Run the gate directly:

cargo test --lib startup_budgets -- --nocapture

The test emits one machine-readable JSON object with top-level language, budgetMs, metrics, and passed fields. Each metric contains samples, medianMs, p95Ms, maxMs, and passed. The fixed protocol is five warmups and 50 measured samples.

This benchmark isolates wrapper overhead. A real Autohand CLI may additionally perform provider authentication, network access, model loading, and other environment-specific readiness work; those live measurements are deliberately reported separately from the deterministic 50 ms gate.